Complex cellular processes such as signal transduction, gene expression, motility and energy metabolism are often implemented using multi-component molecular assemblies. Understanding how these multi-component molecular machines function is an emerging frontier in cell biology, which will begin to define the information gap that exists between our knowledge of the structures of individual proteins and those of cellular organelles. As more networks of interacting proteins emerge from genomics and proteomics, the need for methods to illuminate these potentially disordered complexes will amplify. High resolution electron microscopy is uniquely poised to meet this challenge for a variety of biological specimens that are amenable neither by NMR or X-ray crystallographic techniques. A major focus of my laboratory is the structure determination of large multiprotein complexes by analysis of high resolution images of single molecules. In single particle electron microscopy, images containing large numbers of well-separated protein molecules are recorded using low-dose electron microscopy of frozen-hydrated samples. Individual molecules are computationally selected, sorted into distinct classes, and averaged together to obtain distinct views of the molecule that have a high signal-to-noise ratio. The averaged views are then oriented with respect to each other, and used to reconstruct a model of the three-dimensional structure, which is subsequently improved using refinement algorithms. Using single molecule microscopy, we have defined and interpreted the structure of an icosahedral pyruvate dehydrogenase multienzyme complex, a prototypical example of a multi-step catalytic machine which couples the activity of three component enzymes (E1, E2, and E3) in the oxidative decarboxylation of pyruvate to generate acetyl CoA at the junction of glycolysis and the tricarboxylic acid cycle. The three-dimensional model for a 11 MDa, icosahedral PDH complex, composed of 60 E2 enzymes and 60 E1 enzymes, was obtained by combining a 28 structure derived from electron cryo-microscopy with previously determined atomic coordinates of the individual components of the complex. Analysis of the model provides a number of novel insights into the design and function of this molecular machine. A key feature is that the E1 molecules are located on the periphery in an orientation that allows each of the 60 mobile lipoyl domains tethered to the inner E2 enzyme to access multiple E1 active sites from inside the icosahedral complex. This unanticipated architecture provides a highly efficient mechanism for active site coupling and catalytic rate enhancement, which we propose is achieved by the motion of the lipoyl domain in the restricted annular region between the inner and outer cores of the complex. We are currently refining a second PDH complex comprised of 60 E2 enzymes and 60 E3 enzymes to determine the structural basis of why in vivo the inner icosahedron of 60 E2 molecules is suboptimally occupied with only 48 E1 molecules and 6 E3 molecules typically binding to form the outer protein shell. Analysis of the E1E2 and E2E3 complexes indicate that despite the low occupancy of E3 in the native complex, the lipoyl domains can extend far enough to both mediate active site coupling of E1 and E2 required for the generation of acetyl CoA, and to interact with E3 for the regeneration of an essential disulfide linkage in the lipoyl domain. We are also working actively to identify conditions that lead to outstanding microscopic images, to develop methods to select and accurately align the best molecular images for three-dimensional reconstructions, to reliably interpret these structures, and to develop automated procedures to facilitate the process of obtaining high quality three dimensional models of macromolecular complexes. To this end, we have 1) developed algorithms to collect data automatically on the Tecnai series of electron microscopes, 2) characterized the properties of a 4000 x 4000 pixel digital CCD camera, and assessed the quality of the three-dimensional molecular models constructed from CCD digital images , 3) developed a "core-weighting" method, combined with a grid-threading Monte Carlo approach to enhance the ability to reliably identify the best fit of atomic coordinates of individual components into low resolution maps of larger complexes that are typical of structures determined with the use of single particle electron microscopy and 4) optimized methods for the computational analysis of molecular images . The latter involves methods to accurately orient the molecules, to correct distortions introduced during image collection on the electron microscope, and to enhance the speed of data processing so that it will be possible to analyze the hundreds of thousands of molecular images that will be required to attain near-atomic resolution three-dimensional models of non-symmetrical molecules. We have successfully interfaced our image analysis programs with the Biowulf-Lobos computer cluster to develop parallel computing methodology and a web-based graphical user interface and have used these advances to explore how to improve the resolution of our three-dimensional models. Strategies to achieve the highest resolutions in structures of protein complexes determined by cryo-electron microscopy generally involve averaging information from large numbers of individual molecular images. However, significant limitations are posed by heterogeneity in image quality and in protein conformation that are inherent to large data sets of images. We have demonstrated that the combination of iterative refinement and stringent molecular sorting can be an effective method to obtain substantial improvements in map quality of the 1.8 MDa icosahedral E2 catalytic core. From a starting set of 42,945 images of the core complex, we have shown that using only the best 139 particles in the data set produces a map that is superior to those constructed with greater numbers of images, and that many of the alpha-helices in the structure can be unambiguously visualized in a map constructed from as few as 9 particles. Application of such methods to other macromolecular complexes may greatly facilitate accurate docking of X-ray coordinates of individual component proteins into the density maps obtained by cryo electron microscopy, which in turn, should provide a powerful tool to visualize important macromolecular complexes present in normal and malignant cells.